Anthropic Admits Secret Downgrading of Fable 5 Model for Frontier AI Research
"What they did is, there's a narrow piece where they walk back, which is they said that when it came to things like machine learning, AI research, chip design research, those types of areas, they would kick you to a lesser model but not tell you that. And they would even do things like rewrite your prompt in the background. So they would give you a nerfed answer. They would not tell you what they were doing. They would still charge you for the product that you thought you were getting, and they would never tell you that you were not getting frontier model capability."
About this episode
The All In podcast featuring hosts Jason Calacanis, Chamath Palihapitiya, David Sacks, and David Friedberg opened with extensive debate over Anthropic's controversial Fable 5 model release and AI regulation. The hosts dissected how Anthropic secretly downgrades users conducting AI research to inferior models without disclosure while charging full price, burying the practice in a 319-page document. Simultaneously, CEO Dario Amodei called for a new FDA-style regulatory agency to approve all AI models, prompting accusations of regulatory capture designed to eliminate open-source competitors. Friedberg warned that US restrictions are forcing companies to adopt Chinese open-source models which now outperform American alternatives, creating geopolitical disadvantages. The conversation shifted to Bernie Sanders' proposal for a 50 percent equity seizure from major AI companies to fund a sovereign wealth fund, with Sacks expressing surprising sympathy given AI CEOs' predictions of massive job loss, though he opposes outright confiscation. Friedberg countered that AI is creating productivity gains and hiring booms rather than job losses, citing his own company's aggressive expansion. The episode then pivoted to inflation data showing CPI at 4.2 percent and PPI at 6.5 percent year-over-year, the highest readings since 2022-2023, with hosts attributing the surge to the Iran war and out-of-control government spending. The final segment focused on statistically improbable vote swings in LA's mayoral primary, where Spencer Pratt won in-person voting but saw his late mail-ballot support drop by one-third while opponent Nithya Raman's surged 80 percent. The hosts debated whether California's ballot harvesting laws, automatic ballot mailing, and lack of ID requirements enable systematic fraud or represent legal exploitation, with Sacks declaring he denies the legitimacy of Raman's advancement to the runoff and calling for federal investigation.
Key takeaways
- Anthropic secretly downgrades users conducting AI research to inferior models without disclosure while charging full price, buried in a 319-page document.
- CEO Dario Amodei simultaneously called for new FDA-style regulatory agency to approve all models, prompting regulatory capture accusations designed to eliminate open source.
- Friedberg warned US AI restrictions force American companies to adopt Chinese open source models that now outperform US alternatives, creating geopolitical disadvantages.
- Bernie Sanders proposed 50 percent equity seizure from major AI companies for sovereign wealth fund, gaining surprising sympathy from Sacks given job loss predictions.
- CPI hit 4.2 percent and PPI reached 6.5 percent year-over-year, highest since 2022-2023, attributed to Iran war and government overspending.
- LA mayoral primary showed statistically improbable vote swings with Pratt's late mail ballot support dropping one-third while Raman's surged 80 percent.
- Panel debated whether California's ballot harvesting laws, automatic mailing, and no ID requirements enable systematic fraud or legal exploitation warranting DOJ investigation.